Rotation angle detection method using modulated wave resolver device

Abstract

To provide a method for easily calculating a rotation angle from a detection signal and a device for calculating the rotation angle without the need of an analog circuit and complex detection signal processing for processing the detection signal.SOLUTION: A rotation angle detection method using a modulated wave resolver device includes steps of: generating a reference signal that synchronizes with a synchronous clock and includes N (N is an integer) synchronous clocks in one cycle, and first and second modulated wave driving signals that synchronize with the synchronous clock and are orthogonal to each other; driving first and second input coils of a modulated wave resolver device by using the first and second modulated wave driving signals; acquiring a detection signal from an output coil of the modulated wave resolver device for every half-cycle of the synchronous clock in one cycle of the reference signal and generating modulated wave detection signal data; and generating envelope signal data from the modulated wave detection signal data. The phase of a rotation angle is output from the envelope signal data.SELECTED DRAWING: Figure 4

Description

[Technical field]

[0001] The present invention relates to a method for detecting a rotation angle using a modulated wave resolver device. [Background technology]

[0002] Conventional wound resolvers are composed of an iron core and multiple winding coils, and are electromagnetic absolute displacement detection sensors that detect rotation angles and positions by electromagnetic induction. This type of wound resolver has problems such as large size and weight, and high manufacturing costs. However, since it is composed only of an iron core and winding coils, it has a simple structure, is highly reliable, robust, and has excellent environmental resistance, and is therefore used in a wide range of fields such as aircraft, electric vehicles, robots, and machine tools.

[0003] In addition to the wound-type resolver, modulated wave resolver devices disclosed in Patent Documents 1 and 2 are also known. Compared to conventional wound-type resolvers, modulated wave resolvers can use high-frequency drive signals, which increases detection sensitivity and allows for a significant reduction in the number of coil turns. For this reason, in modulated wave resolver devices, the input and output coils included in the rotating mechanism are patterned and formed on a printed circuit board. This gives them the excellent characteristics of being small, lightweight, and low manufacturing costs. Moreover, Patent Document 2 discloses a modulated wave resolver device using a high-frequency (approximately 500 kHz) modulated wave and a method for detecting a rotation angle by an analog control method.

[0004] Known resolvers are of two-phase excitation, one-output type and one-phase excitation, two-output type. This example illustrates application to a two-phase excitation, one-output modulated wave resolver. This type of resolver has the advantage that signal processing and control of the rotation angle on the output side are easy because it has one output. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Patent No. 3047231 [Patent Document 2] Patent No. 6811369 Summary of the Invention [Problem to be solved by the invention]

[0006] In conventional resolver devices, in order to calculate the rotation angle from the detection signal from the resolver's output coil, the detection signal was processed via a synchronous detection circuit, then passed through a low-pass filter to remove the modulated wave, only the envelope signal was extracted, and further, the zero potential point of the envelope signal was detected to calculate the rotation angle. This required complex signal processing using analog circuits.

[0007] What is needed is a method for easily calculating the rotation angle from a detection signal without using an analog circuit for processing the detection signal or performing complex detection signal processing, and a method for calculating the rotation angle. The object of the present invention is to provide an apparatus for performing the above-mentioned [Means for solving the problem]

[0008] A method for detecting a rotation angle according to the present invention is a method for detecting a rotation angle by a modulated wave resolver device, and includes the steps of: generating a reference signal that is synchronized with a synchronous clock and includes N synchronous clocks (N is an integer) per one period; generating a first modulated wave drive signal and a second modulated wave drive signal that are synchronized with the synchronous clock and are orthogonal to each other; driving a first input coil and a second input coil of the modulated wave resolver device with the first modulated wave drive signal and the second modulated wave drive signal; acquiring a detection signal from an output coil of the modulated wave resolver device every half period of the synchronous clock during one period of the reference signal; The method includes the steps of: generating 2N pieces of modulated wave detection signal data excluding DC components contained in the signal; generating a modulated wave detection signal data group including the modulated wave detection signal data; inverting the sign of even-numbered data among the modulated wave detection signal data to generate sign-inverted modulated wave detection signal data; and combining the odd-numbered data of the modulated wave detection signal data group with the sign-inverted modulated wave detection signal data to generate an envelope signal data group including 2N pieces of envelope signal data.The method is characterized in that the rotation angle phase θj is calculated from the 2N pieces of envelope signal data using equations (1), (2) and (3) and output for each period of the reference signal.

number

[0009] In the rotation angle detection method according to the present invention, the rotation angle phase is calculated from 2N envelope signal data using equations (1), (2), and (3), so that the rotation angle phase θj can be determined with high accuracy despite various factors that disrupt accuracy, such as voltage fluctuation, drift, noise, distortion, and variation in the modulated wave.

[0010] In addition, in the rotation angle detection method according to the present invention, in the step of generating the modulated wave detection signal data group, the modulated wave detection signal data group may include 2N×n pieces of modulated wave detection signal data acquired during n periods (n is an integer equal to or greater than 2) of the reference signal, and the rotation angle phase may be output for each of the n periods. Since this detection method uses a long 2N×n envelope signal data sequence, it is possible to effectively deal with various factors that disrupt accuracy, such as voltage fluctuations, noise, distortion, and variations in the modulated wave, at an even higher level, and to determine the rotation angle phase θj with high accuracy.

[0011] In addition, in the method for detecting a rotation angle according to the present invention, the method further comprises the steps of: storing the envelope signal data group generated during a first period of the reference signal as a first envelope signal data group in a storage device; a second period of the reference signal is set at a timing shifted by a first synchronization clock included in the first period, the 2N-1th and 2Nth modulated wave detection signal data are acquired from the 2N modulated wave detection signal data included in the second period, extracted envelope signal data is generated from the acquired modulated wave detection signal data, and the extracted envelope signal data is synthesized with the 3rd to 2Nth envelope signal data from the stored first envelope signal data group, generating a second envelope signal data set including 2N envelope signal data; The rotation angle phase is calculated from 2N pieces of the second envelope signal data group. In this detection method, the first envelope signal data group stored in the storage device is used as the second envelope signal data group, so that the rotation angle phase θj can be output at high speed in synchronization with the synchronous clock with a small amount of calculation.

[0012] In addition, in the rotation angle detection method of the present invention, the step of generating the modulated wave detection signal data may be characterized in that even-numbered data are deleted from the 2N detection signals, and the modulated wave detection signal data is generated from odd-numbered data.

[0013] In addition, the method for detecting a rotation angle according to the present invention may further include a step of generating a detection clock that is delayed by a predetermined time from the synchronous clock. The step of acquiring the detection signal may be characterized in that the detection signal is acquired in synchronization with the detection clock.

[0014] Furthermore, in the rotation angle detection method of the present invention, the rotation angle phase θj output for each period of the reference signal may be sequentially stored in a storage device, and two adjacent rotation angle phases θj−1, and θj from the stored rotation angle phases are read out, and a corrected rotation angle θcj is calculated by the following equation (4). θcj=θpj-sin{2(θpj-k1)π / 180}·k2+k3 (4) Here, θpj, k1, k2, and k3 can be defined by the following equations. θpj=θj+(θj-θj-1) / 2 k1=(θj-θj-1)·α1 k2=(θj-θj-1)·α2 k3=(θj-θj-1) · (θj-θj-1) · α3 Here, α1, α2, and α3 are constants. Effect of the Invention

[0015] In the rotation angle detection method according to the present invention, the rotation angle phase θ is calculated from 2N envelope signal data series of the modulated wave using equations (1), (2), and (3), so that it is possible to deal with various factors that disturb the accuracy such as voltage fluctuation, drift, distortion, and variation of the output modulated wave, and to calculate the rotation angle phase θ with high accuracy. Since a numerical calculation is used as a means for calculating the rotation angle phase θ, conventional analog circuits and complicated detection signal processing are not required. [Brief description of the drawings]

[0016] [Figure 1] FIG. 1 is a diagram for explaining the basic configuration of a modulated wave resolver. [Diagram 2]FIG. 2 is a diagram showing a configuration in which a stator and a rotor of a modulated wave resolver are combined. [Diagram 3] FIG. 3 is a diagram showing a modulation method of a modulated wave resolver. [Figure 4] FIG. 4 is a diagram showing the configuration of a modulated wave resolver device incorporating a machine learning algorithm. [Diagram 5] FIG. 5 is a diagram showing the configuration of a modulated wave resolver device that outputs signals A and B to an external controller. [Figure 6] FIG. 6 is a diagram for explaining the sampling timing of the received modulated wave. [Figure 7] FIG. 7 is a diagram showing the acquisition of received modulated wave data. [Figure 8] FIG. 8 is a diagram showing the principle of detecting the rotation angle phase θ by machine learning. [Figure 9] FIG. 9 is a diagram for explaining the envelope signal data sequence {yi} for one period of the received signal when the rotor is rotating. [Figure 10] FIG. 10 is a diagram showing factors and configuration for determining the rotor rotation angle. [Figure 11] FIG. 11 is a diagram showing an example of adjacent error at a rotation speed of 6000 rpm. [Figure 12] FIG. 12 is a diagram showing an example of an uncorrected error waveform at a rotation speed of 6000 rpm. [Figure 13] FIG. 13 is a diagram showing an example of the rotation angle error after correction at a rotation speed of 6000 rpm. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0017] (First embodiment) Hereinafter, a first embodiment of the present invention will be described with reference to FIGS. Before describing the control method of the modulated wave resolver according to the present invention, the operating principle of the current modulated wave resolver device will be described.

[0018] Figure 1 shows the principle configuration diagram of a two-phase excitation, one-output type modulated wave resolver. This type of modulated wave resolver operates on a simple principle of driving (exciting) the input coil of the resolver mechanism with a modulated wave (carrier wave) whose envelope contains orthogonal SIN and COS signal waves, and measuring the rotation angle or rotation speed by detecting the rotation angle phase θ of the modulated wave that is output from the output coil in accordance with the rotation angle of the resolver. Note that in the following explanation, the rotation angle phase may be expressed as θj (j is an integer).

[0019] Figure 2 shows the combined configuration of the stator and rotating rotor of a modulated wave resolver. The stator and rotating rotor face each other with a gap of less than 1 mm, and their respective coil patterns are arranged on the inside. The SIN input coil, COS input coil, and output coil shown in Figure 1 are arranged on the printed circuit board on the stator side. In addition, a pattern that generates a rotating magnetic field is arranged on the printed circuit board on the rotating rotor side. As the rotor rotates, the rotation angle phase θ of the output modulated wave is detected from the output coil of the fixed stator.

[0020] FIG. 3 is a diagram for explaining a modulation method of a modulated wave resolver. The signal wave is synchronized with a reference signal (period T) that defines the basic operation of the resolver. In current modulated wave resolvers, the signal wave is driven (carried) by a modulated wave, i.e., an AM modulated carrier wave. Therefore, the signal wave is expressed as the envelope of a SIN and COS modulated wave.

[0021] As an example of an AM modulated wave, the following shows one period of a reference signal, or 32 waves x 2 = 64 waves. Synchronized with a half period of the reference signal, the first half of the modulated wave shows positive polarity, and the second half shows negative polarity. The phase of the modulated wave is shifted by 180° between positive and negative polarities. Polarity inversion is performed continuously in synchronization with a half period of the reference signal.

[0022] Figure 4 shows the configuration of a modulated wave resolver device incorporating a machine learning algorithm. The modulated wave resolver device 1 has a controller A-2 configured with programmable logic such as a DSP or FPGA, which manages and controls a resolver mechanical mechanism 5 and detects the rotation angle phase θ. The modulated wave resolver device 1 is composed of a controller A-2 incorporating a signal generating unit, a first modulated wave drive circuit 3 which is a drive circuit that drives (excites) the first input coil 10, a second modulated wave drive circuit 4 which is a drive circuit that drives (excites) the second input coil 11, a resolver mechanical mechanism 5, a differential amplifier 6 which detects the rotation angle phase θ from the output coil 12, an anti-alias filter 7, and an AD converter 8. The interface circuit 9 is a circuit for outputting rotation angle information (corrected rotation angle phase θ) to an external device.

[0023] In the present invention, a modulated wave synchronized with a reference signal (period T) and a synchronous clock is used as an excitation circuit for driving the first input coil 10 and the second input coil 11 of the resolver. In the SIN series, a modulated wave is output from the first modulated wave excitation circuit 3 of the resolver input coil excitation unit in synchronization with the synchronous clock by the first modulated wave excitation signal, thereby exciting the resolver first input coil 10. In the COS series, a modulated wave is output from the second modulated wave excitation circuit 4 of the resolver input coil excitation unit in synchronization with the synchronous clock by the second modulated wave excitation signal, thereby exciting the resolver second input coil 11. The SIN series and COS series signals are mutually orthogonal.

[0024] As the resolver mechanism 5 rotates, a rotation angle signal is output from the output coil 12 in synchronization with the synchronous clock. The rotation angle phase information of the output modulated wave is passed through a differential amplifier 6 that detects rotation angle phase information, an anti-aliasing filter 7, and an AD converter 8 to a controller. The signal is input to the rotation angle detector of A-2. The data is discretized by the AD converter 8 through the anti-aliasing filter 7, which follows the sampling theorem. The anti-aliasing filter 7 is made up of a high-order low-pass filter that removes frequency components higher than the modulation wave frequency (synchronized with the synchronous clock), which is the sampling period of the AD converter 8, and extracts only the fundamental component of the modulation wave.

[0025] The rotation angle calculation unit of controller A-2 is equipped with a CPU, RAM memory, and non-volatile program within its calculation device, and executes product-sum calculation processing of a rotation angle detection calculation algorithm using machine learning, which will be described later, and is able to calculate the rotation angle phase θ with high accuracy from signals A and B. In this case, a high-precision trigonometric function calculator for calculating the rotation angle phase θ is required within the controller A-2.

[0026] In the modulated wave resolver device in FIG. 4, the rotation angle phase θ is calculated and corrected in the controller A-2. However, there is also a method in which the signals A and B from the rotation angle calculation unit of the controller A-2 are output to an external controller, and the external controller 13 performs calculation processing and calculates and corrects the rotation angle phase θ. FIG. 5 shows a configuration diagram of a modulated wave resolver device that outputs signals A and B to the external controller B. Upon receiving signals A and B from the calculation unit in the controller B-14, the external controller 13 calculates and corrects the rotation angle phase θ externally. With this configuration, it is possible to eliminate the need for a high-precision trigonometric function calculator that calculates the rotation angle phase θ in the controller B-14, and to configure a low-function and inexpensive controller B-14. On the other hand, the external controller 13 uses a high-performance system master microcomputer that has advanced functions such as a high-speed serial line for calculating the rotation angle phase θ and a high-precision trigonometric function calculator.

[0027] FIG. 6 is a diagram for explaining the sampling timing of the received modulated wave. An AD converter 8 is used to sample and obtain the received modulated wave data. In order to accurately read the received modulated wave data, sample and hold (S&H) is performed at the peaks of the received modulated wave, and this data is read by the AD converter 8. The received modulated wave from the resolver is output with a delay from the synchronous clock due to factors such as resolver wiring delays. For this reason, a detection clock that is phase-delayed by a specified time from the synchronous clock is generated separately and used for the S&H (sample and hold) signal. In synchronization with the detection clock, the AD converter 8 alternately samples the odd and even peaks of the received modulated wave, and modulated wave detection signal data for the odd and even peaks is obtained.

[0028] Figure 7 shows how to obtain received modulated wave data. The received modulated wave that has passed through the anti-aliasing filter 7 is synchronized with the detection clock, and its envelope is synchronized with the reference signal period (T). According to the principle of the rotation angle of the modulated wave resolver, the rotation angle phase θ of the received modulated wave moves based on the period T of the reference signal, following the rotor rotation of the resolver. By inverting the sign of the even-numbered data in the modulated wave detection signal data at the center line of the modulated wave amplitude and combining it with the odd-numbered data, a series of 2N envelope signal data (period T) is obtained that has twice the frequency of the modulated wave and has a sine-wave envelope during the reference signal period (T).

[0029] Another method for obtaining a sine-wave envelope signal data series is to delete all even-numbered data that have sign inversions, but this reduces the number of data by half and reduces the rotation angle detection accuracy.

[0030] If the conventional analog θ phase detection method could be realized using a new digital method, many of the issues encountered with analog circuits would be resolved and the accuracy of rotation angle detection would be improved. From FIG. 1, the principle of the resolver rotation angle is based on the following trigonometric addition theorem.

number

[0031] Let us look at the output side of the principle equation of the resolver that detects the rotation angle. In the principle equation, the output rotation phase θ from the resolver is expressed as an ideal envelope signal wave {sin(ωt±θ)}, with a modulated wave synchronized with the detection clock as the carrier wave. The actual resolver output data corresponding to the signal wave {sin(ωt±θ)} corresponds to a 2N-piece envelope signal data series with a sine wave-shaped envelope obtained by sign-inverting the even-numbered data obtained in the process shown in Figure 7.

[0032] The output modulated wave data, which has a sine wave envelope, contains a variety of distortions and variations that disrupt accuracy, such as voltage fluctuations, drift, noise, distortion, and variation, due to a combination of factors such as manufacturing distortion of the resolver pattern, mechanical rotation angle error between the stator and rotor, and electrical error in resolver excitation.

[0033] In the present invention, we have conceived of a method of mapping an ideal principle equation of a resolver onto 2N pieces of sine wave envelope signal data so that the error of the resolver rotation angle phase θ is minimized, and reasonably estimating the rotation angle phase θ mathematically and statistically. As a means for this, we will exemplify a method of estimating θ using machine learning (AI).

[0034] FIG. 8 is a diagram for explaining the principle of detecting the rotation angle θ by machine learning. The theoretical equation showing the actual rotation angle of a modulated wave resolver can be expressed as the following sine function by adding the parameters of amplitude A and offset ε to the original theoretical equation, where t is the time. A sin(ωt±θ)+ε The gist of the present invention is that, as a means for determining the rotation angle phase θ, the acquired sinusoidal envelope signal data series of the resolver is mapped to the sine function of the resolver so as to minimize the error, thereby determining the optimal rotation angle phase θ.

[0035] The actually acquired received modulated wave data contains distortion and variation, and in addition, the parameter to be obtained is not only the rotation angle phase θ, but also the additional parameters amplitude A and offset ε, which must be obtained at the same time. In the present invention, the acquired modulated wave data with distortion and variation is mapped to a sine function of the ideal principle equation of the resolver so that the squared error is minimized, and a least squares learning model of machine learning (AI) is applied to obtain the optimal rotation angle phase θ. At this time, what is actually desired to be obtained is only the rotation angle phase θ, and the amplitude A and offset ε are additional, or in other words, unnecessary parameters.

[0036] In the least squares learning model, the squared error between the model output f{A,θ(xi),ε} and the training output {yi} The rotation angle phase θ is obtained through learning so as to minimize L(A, θ, ε). The basis function in this case is shown in equation (5). Basis function: yi=Asin(ωxi+θ)+ε (5) The squared error L(A, θ, ε) is expressed by the following equation (6).

number

[0037] To minimize the training squared error L(A,θ,ε), The following three partial differential equations must be satisfied simultaneously: (∂ / ∂θ){L(A,θ,ε)}=0 (7) (∂ / ∂A){L(A,θ,ε)}=0 (8) (∂ / ∂ε){L(A,θ,ε)}=0 (9) Since equations (7) to (9) are nonlinear, they cannot be found analytically, and the parameters A, θ, and ε are found by using an iterative calculation algorithm such as a successive approximation method. For this reason, in the past, the calculation steps and processing time were long, and it was not possible to implement an algorithm for continuously finding the rotation angle phase θ within a practical time frame in the device.

[0038] To overcome this situation, we first focus on equation (7) which calculates θ. From equations (5) and (6),

number

[0039] In equation (10), if we limit the sigma accumulation range to the orbit (i = 1 to n2N), we get With ω=2π / 2N and xi=i, we obtain the following result.

number

number

[0040] From equation (13), θ can be analytically solved by the following method. cos (ωxi + θ)= cos(ωxi) cos θ ? sin(ωxi) sin θ that's why, yi cos (ωxi + θ) = yi cos(ωxi) cos θ ? yi sin(ωxi) sin θ When we take the sum, we get the following equation (14).

number

number

number

[0041] The calculation formula (17), in which the calculation range of sigma is i=1 to 2N (or n2N), increments i in synchronization with half the detection clock, sequentially acquires input data yi, and accumulates A and B using product-sum calculations, allowing a series of processes to be executed in real time (at this time, the even data undergoes sign inversion processing when input). In other words, 2N envelope signal data series yi, each with a sine wave envelope, can be sequentially input in real time and accumulated. By pairing the odd data and even (sign inversion) data of i, it becomes possible to accumulate A and B in synchronization with the detection clock. In this case, clock synchronization makes it easy to carry out subsequent continuous processing. When the integration process is completed, the rotation angle phase θ can be obtained from A and B at that point in time. Based on a request from the external interface circuit 9, the rotation angle phase θ is output asynchronously.

[0042] The essence of this invention is that by limiting the range of acquired data to revolutions in equation (14), it is possible to eliminate the parameters of amplitude A and offset ε, resulting in a simple conditional equation consisting only of the rotation angle phase θ. By using simple product-sum calculations, it is possible to execute the calculations in real time using inexpensive multipliers. Furthermore, for COS and SIN calculations, function tables can be prepared in non-volatile memory, simplifying the calculations. As a result, it is possible to incorporate and implement the algorithm in the rotation sensor controller at a practical price.

[0043] High-precision analog-controlled rotation sensors with a rotation angle error of ±0.01° or better (equivalent to 16-bit resolution) are sensitive to minute changes in the control circuit, such as voltage fluctuations, temperature changes, and aging, which can cause rotation angle errors and are extremely difficult to address. The machine learning algorithm of the present invention has the excellent property of being able to theoretically eliminate the effects of signal amplitude A and offset ε, and stably detect only the rotation angle θ from a floating input signal that includes disturbances.

[0044] In machine learning, the more data there is, the more accurate the rotation angle can be. In the method of the present invention, the amount of data can be scalably changed (n2N) in units of one cycle of the reference signal (2N), and the calculation range can be changed without changing the hardware.

[0045] A typical modulated wave resolver using an analog control method has a reference signal frequency of 7 kHz and a modulated wave frequency of about 450 kHz. Since the rotation angle detection timing is the zero cross point of the signal wave, the rotation angle detection period is also limited to the reference signal frequency of 7 kHz. In contrast, in this method, if the rotation angle detection cycle is set to a synchronous clock (for example, 450 kHz) in the following procedure, the rotation angle phase θ can be output continuously at high speed.

[0046] Here, for simplicity, the envelope signal data range is limited to one period (2N), but it can be expanded to n2N. In machine learning, accuracy can be improved by increasing the amount of data and using big data. It has been demonstrated that by incorporating this expansion function expanded to n2N into an actual prototype, accuracy can be improved without increasing the amount of calculations.

[0047] First, prepare a cyclic memory with 2N addresses to store the envelope signal data group data yi cos{(2π / 2N)×i} and yi sin{(2π / 2N)×i} that make up A and B. A and B are obtained by adding and accumulating the data of the envelope signal data groups in this cyclic memory. During the first period, the envelope signal data group is filled into the circular memory, and the data is added and accumulated. A and B for one period are calculated. From the second period onwards, the address of the cyclic memory is updated in synchronization with the synchronous clock, and the old envelope signal data for two addresses is obtained from the updated address. Next, the new envelope signal data for two addresses is written onto the old envelope signal data at the update address to update it. The new A and B are obtained by correcting the difference between the old individual envelope signal data and the new envelope signal data in the current A and B, respectively.

[0048] With this detection method, a group of envelope signal data is stored in a cyclic memory, and the rotation angle phase θ can be output at high speed in synchronization with a synchronous clock with a small amount of calculation and without being restricted to output in units of one period.

[0049] In conventional analog control methods, θ was calculated from a counter circuit. There was an upper limit to the clock frequency that drives the counter, and furthermore, there was jitter in the synchronous detection circuit, so there was an upper limit to the numerical value that expressed θ. This upper limit also limited the number of resolutions of the resolver. In contrast, in the method of the present invention, the rotation angle phase θ can be expressed in single-precision or double-precision floating point as a numerical value resulting from a numerical calculation based on product-sum calculations. Therefore, there is an advantage that there is no unstable jitter like in analog control methods.

[0050] In the method of the present invention, all the computational processing of the machine learning algorithm can be configured with a combination of existing inexpensive commercially available parts such as AD converters, FPGAs, and DSPs (Digital signal processors). Furthermore, since the configuration is simple, it is easy to make it into an ASIC. In the present invention, the ADC and subsequent parts are configured with numerical calculations based on the machine learning algorithm. With the evolution of semiconductor integration, we can enjoy the benefits of simultaneous progress in speed, performance, and cost reduction.

[0051] Second Embodiment In the first embodiment, a detailed description was given of "driving (exciting) the input coil of the resolver mechanical mechanism with a modulated wave (carrier wave) having orthogonal SIN and COS signal waves in its envelope, and detecting the rotation angle phase θ of the modulated wave output in accordance with the rotation angle of the resolver from the output coil, thereby measuring the rotation angle or rotation speed." With this detection method, it is possible to accurately detect the rotation angle when the rotor rotation speed of the resolver is relatively low, but it has been found that as the rotation speed increases, errors occur depending on the rotation speed. Next, the cause of this error and a method for solving it will be described in detail.

[0052] In the first embodiment, the amplitude A and offset ε can be eliminated by limiting the range of the acquired data to a revolution, and the above formula (14) can be expressed simply, and a method for determining the rotation angle phase θ from the envelope signal data series {yi} from the output coil of the resolver has been described. In the second embodiment, the accumulation range of sigma is also limited to a revolution (i = 1 to n2N), so formulas (15), (16), and (17) can be used. However, as described later, when the rotor rotates at a high speed, the acquired data frequency ω changes. For this reason, a rotation angle error occurs. Below, the cause of the rotation angle error and the method for correcting the rotation angle error will be described in detail.

[0053] Figure 9 shows the envelope signal data sequence {yi} for one period (period T) of the received signal when the rotor is rotating. Figure 9(a) shows the state of the extremely low-speed rotation signal wave. The signal wave period of the received signal (envelope signal data sequence {yi}) when the rotor is rotating at an extremely low speed including when it is stationary matches one period (period T) of the received signal. In this state, the rotation angle phase θ can be calculated with high accuracy using the envelope signal data sequence {yi} for one period of the received signal and equations (16) and (17) shown in the first embodiment.

[0054] Figures 9(b) and (c) show the behavior of the high-speed rotation signal wave. When the rotor rotates at high speed, the signal wave period (envelope signal data series {yi}) of high speed rotation no longer matches one period (period T) of the received signal. This phenomenon will be described later, but depending on the direction of rotation of the rotor, the received signal wave period may be shorter (Figure 9(b)) or longer (Figure 9(c)) than one period (period T) of the received signal. In such cases, when detecting the rotation angle using the envelope signal data series {yi} for one period (period T) of the received signal, an error will occur in the rotation angle detection depending on the rotation speed due to the difference between the reception period T and the period of the envelope signal data series {yi}, which is the signal. This phenomenon is caused by the difference in the received data frequency of the envelope signal data series {yi} depending on the rotation speed during the reception cycle T.

[0055] The envelope signal data sequence {yi} for one period of the excitation wave is defined by the following resolver principle equation: yi=sin{ωt±θ} Since θ corresponds to the rotational angle phase of the rotor, when the rotor rotates, the above resolver principle equation can be expressed by the following equation. θ=ωr×t=2πFt Therefore, the above resolver principle equation can be expressed as the following equation. yi=sin{ωt±θ} =sin{ωt±ωr×t} =sin{(ω±ωr)×t} (18) =sin{2πft±2πFt} =sin{2πt(f±F)} =sin{2πt×f (1±F / f)} (19) Where: ωr: Angular frequency of rotor rotation f: Excitation wave frequency Hz (f=1 / T) F: Rotor rotation frequency Hz (rpm / 60) ± indicates the direction of rotor rotation.

[0056] Equation (18) can be interpreted as the combination of ω (electrical system) and ωr (mechanical system) into one composite wave with frequency (f+F). Equation (19) shows that the frequency of the data series {yi} expands and contracts in response to changes in the rotor speed F (Hz). This phenomenon causes a rotation angle error, which will be explained later.

[0057] Next, the method for correcting the rotation angle error described above will be described in detail. In formula (19), the following conditions (1) and (2) are satisfied. (1) When the rotor is stationary (F=0), the acquisition period of the data series {yi} coincides with the excitation wave period (T=1 / f). Also, the frequency of the data series {yi} during this period is f. It has become. (2) When the rotor is rotating, the acquisition period T of the data series {yi} does not change as 1 / f. However, the frequency of the data series {yi} during this period is affected by the rotational frequency of the rotor according to equation (19) and changes as f(1±F / f).

[0058] In (2), the frequency of the data series {yi} changes by ±F / f due to the rotor rotation speed F. Furthermore, during the acquisition period T of the data series {yi}, the rotor rotates by ±360°×F / f. If the rotation angle of the rotor during this period is Δθ, then we obtain the following equation (20). Δθ=±360°×F / f (20)

[0059] FIG. 10 shows the factors and configuration for determining the rotor rotation angle. The data series {yi} waveform is the input envelope data series {yi} for one period of the excitation wave when the rotor is rotating, and is indicated by arrows j-1 and j along with the direction of rotation. As an example, the explanation will proceed assuming that the rotor rotation direction is forward rotation (counterclockwise). For reverse rotation (forward clockwise), the arrow direction is reversed. In addition, the rotation angle phase θ calculated from equations (16) and (17) using the {yi} data series is referred to as the AI ​​rotation angle value (θj-1, θj).

[0060] As the rotor rotates, the {yi} data series starts from the 0° absolute origin. The average angle Δθ of the {yi} data series over the rotor rotation angle of 360° is again given by equation (20). Δθ=360°×F / f (20) In reality, the rotor rotation is accelerated and decelerated, and the rotation speed is not constant, but within the small interval Δθ, the change in the rotation speed is considered to be small and linear. The received signal {yi} data series is acquired as data for the Δθ interval, and the AI ​​rotation angle value θj calculated from this {yi} data series indicates the intermediate position of Δθ. From this, the intermediate position of Δθ is defined as the rotation angle phase θj.

[0061] Furthermore, Δθ is expressed as the difference between the AI ​​rotation angle values ​​of adjacent {yi} data series, which are θj and θj−1. Δθ=(θj―θj-1) (21) Since the sampling period of the {yi} data series is constant, the difference Δθ calculated from equation (21) also indicates the rotational speed of the rotor.

[0062] FIG. 11 shows an example of adjacent difference obtained by simulation at a rotation speed of 6000 rpm. The adjacent difference Δθ is not a constant value, but has a slight oscillation of ±0.0728° between adjacent data with a center value of 5.14285°. The oscillation frequency has two peaks within 360° of rotor rotation.

[0063] The pre-correction rotation angle to be obtained is indicated by the rotor rotation angle θpj of the leading end value of the final data θj of the {yi} data set. The rotation angle θpj is expressed by the following equation (22) using θj and θj-1. θpj=θj+(θj―θj-1) / 2 (22) Here, when the rotation direction is reversed, θj-1>θj. From equation (22), it can be seen that the rotation angle precision of the rotation angle θpj is closely related to the adjacent angle difference (θj-θj-1), which is also an index of the rotation speed.

[0064] FIG. 12 shows an example of an uncorrected rotation angle error curve at a rotation speed of 6000 RPM. The rotation angle error waveform of the rotation angle θpj=θj+(θj−θj−1) / 2 obtained from the simulation results is a quadratic sine curve with an offset (having two peaks within one period of 360°) as shown in FIG. The error correction of the rotation angle is realized by correcting the error angle of the rotation angle θ using a function according to the rotation speed F of the rotor. The correction term can be realized by setting a function that cancels this error curve.

[0065] The correction of the rotation angle is required to cover a predetermined speed range using a common formula. The main point of this invention is that the speed of the rotation angle (θj-θj-1) is introduced as a common parameter to correct the current rotation angle θpj. As shown in Fig. 10, there is a slight vibration in (θj-θj-1), which also becomes an error factor, but this is also included in the coefficients k1 for the phase of the SIN function, k2 for the amplitude, and k3 for the offset. The corrected rotation angle θcj is defined by the following equation, including correction terms for correcting the phase, amplitude, and offset in the current rotation angle θpj. θcj=θpj - correction term (23) Correction term={2 sin{ (θpj-k1) π / 180}*k2+k3} (24) k1 = (θj - θj-1) * α1 k2 = (θj - θj-1) * α2 k3 = (θj-θj-1)^2*α3 Here, the constants α1, α2, and α3 are set to optimal values ​​within a predetermined range of rotation speeds. Furthermore, by finely adjusting the constants α1, α2, and α3 depending on conditions such as the direction of rotation, the accuracy can be further improved.

[0066] FIG. 13 shows an example of the rotation angle error after correction at a rotation speed of 6000 rpm. This shows the relationship between the rotation angle error corrected using the optimally calculated α1, α2, and α3 under a simulation environment. At a rotation speed of 6000 RPM, the second-order error of the SIN function is completely corrected, and only a minute fourth-order residual error with an amplitude of about ±0.0043° remains. Compared to the error before correction of ±0.4° shown in FIG. 12, the error has been significantly improved to about 1 / 100, demonstrating the effect of the present invention.

[0067] Below is an example of the residual error when common correction constants α1, α2, and α3 are used to correct for different rotation speeds. It can be seen that the correction method of the present invention, which uses the difference (θj-θj-1) as a common parameter, is effective in being applicable to a wide range of rotation speeds. Rotational speed Residual error 1000rpm ±0.0005° 3000rpm ±0.0022° 6000rpm ±0.0043° 12000rpm ±0.0390°

[0068] The embodiment described above is merely an example, and various modifications of the present invention are possible without departing from the scope of the claims. For example, in the description of this example, an example is shown in which the rotation angle phase θ is obtained from the modulated wave output in a two-phase excitation, one-output system, but it is also possible to extend the example to a one-phase excitation, two-output system, and the application of this example in which the rotation angle phase θ is obtained from the modulated wave output is within the scope of the present invention. The various numerical values ​​in the examples used for the purpose of explanation are standard examples, and it goes without saying that in the actual exercise of a patent, they will be changed depending on various conditions and factors. [Explanation of symbols]

[0069] 1. Modulated wave resolver device 2 Controller A 3. First modulated wave driver circuit 4. Second modulated wave driver circuit 5 Resolver mechanism 6 Differential Amplifier 7 Anti-aliasing Filter 8. A / D Converter 9 Interface Circuit 10 First input coil 11 Second input coil 12 Output coil 13 External Controller 14 Controller B

Claims

1. A method for detecting a rotation angle by a modulated wave resolver device, comprising: A step of generating a reference signal synchronized with a synchronous clock and including N synchronous clocks (N is an integer) per period, a first modulated wave drive signal synchronized with the synchronous clock, and a second modulated wave drive signal which are orthogonal to each other; driving a first input coil and a second input coil of the modulated wave resolver device by the first modulated wave drive signal and the second modulated wave drive signal; A step of acquiring a detection signal from an output coil of the modulated wave resolver device every half cycle of the synchronous clock during one cycle of the reference signal, and generating 2N pieces of modulated wave detection signal data excluding DC components contained in the detection signal and a modulated wave detection signal data group including the modulated wave detection signal data; a step of inverting the sign of even-numbered data of the modulated wave detection signal data to generate sign-inverted modulated wave detection signal data; a step of synthesizing the odd-numbered data of the modulated wave detection signal data group and the sign-inverted modulated wave detection signal data to generate an envelope signal data group including 2N envelope signal data, A method for detecting a rotation angle, comprising: calculating a rotation angle phase θj from 2N pieces of envelope signal data using equations (1), (2) and (3) and outputting the calculated phase θj for each period of the reference signal. [0010] Here, yi (i=1 to 2N) represents envelope signal data, and θj (j is an integer) represents the rotation angle phase.

2. In the step of generating the modulated wave detection signal data group, the modulated wave detection signal data group includes 2N×n pieces of modulated wave detection signal data acquired during n periods (n is an integer equal to or greater than 2) of the detection signal, 2. The method for detecting a rotation angle according to claim 1, wherein the rotation angle phase is output every n periods.

3. storing the set of envelope signal data generated during a first period of the detection signal as a first set of envelope signal data in a storage device; the step of setting a second period of the detection signal at a timing shifted by a first synchronization clock included in the first period, acquiring 2N-1st and 2Nth modulated wave detection signal data out of 2N pieces of modulated wave detection signal data included in the second period, generating extracted envelope signal data from the acquired modulated wave detection signal data, and synthesizing the extracted envelope signal data with the 3rd to 2Nth envelope signal data out of the stored first envelope signal data group to generate a second envelope signal data group including 2N pieces of envelope signal data, 2. The method for detecting a rotation angle according to claim 1, further comprising the step of calculating a rotation angle phase from the group of 2N second envelope signal data.

4. 2. The method for detecting a rotation angle according to claim 1, characterized in that, in the step of generating the modulated wave detection signal data group, even-numbered data are deleted from the 2N detection signals, and the modulated wave detection signal data group is generated from odd-numbered data.

5. generating a detection clock delayed by a predetermined time from the synchronous clock; 2. The method for detecting a rotation angle according to claim 1, wherein, in the step of acquiring the detection signal, the detection signal is acquired in synchronization with the detection clock.

6. The method for detecting a rotation angle according to claim 1, characterized in that the rotation angle phase θj output for each period of the reference signal is stored in a sequential storage device, and two adjacent rotation angle phases θj-1 and θj are read out from the stored rotation angle phases, and a corrected rotation angle θcj is calculated by equation (4). θcj=θpj-sin{2(θpj-k1)π / 180}・k2+k3 (4) Here, θpj, k1, k2, and k3 can be defined by the following equations. θpj=θj+(θj-θj-1) / 2 k1=(θj-θj-1)・α1 k2=(θj-θj-1)・α2 k3=(θj-θj-1)・(θj-θj-1)・α3 Here, α1, α2, and α3 are constants.

Images